buremba

damn, my password is here

Is there any tool that does the other way around? I simply need an alternative to cloudflared tunnel (https://blog.cloudflare.com/tunnel-for-everyone/) for exposing localhost port to a public domain that lets me supports anonymous clients. All cloud solutions charge based on users so they unfortunately doesn’t work

It would also be pretty interesting to see a modular system where we can run Python scripts as well

Observable also handles the execution order for you.

I think this would be great as an evolution of MDX (https://mdxjs.com/). MDX is already pretty popular for documentation and it plays well with React but unfortunately there is no framework that adds interactivity to MDX which will enable use cases like data applications.

Do you support scaling to zero? I wonder if native offerings of cloud providers (Cloud SQL/Alloydb or Aurora) still make sense as keeping hundreds of PG instances at scale will likely be a challenge if you're managing them from your control plane.

Also, is there any compliance that requires it to be in different Postgresql servers? I assume most companies just use some sort of isolation (tenant_id column or dedicated tenant database/table) so I wonder if this problem could better be solved as a proxy layer.

SyncLite looks great for companies that are starting to build from scratch. Ingest data in an embedded database and via CDC move the data into a centralized database. I'm also trying out a similar idea but only it's a bit reverse approach, from Cloud Warehouse -> Embedded DuckDB to reduce the compute cost for BI and embedded analytics use-cases. The combination of cloud and embedded databases is the future IMO.

For the project I'm working on, the tech such as Apache Iceberg and embedded DuckDB enables querying Snowflake + BigQuery tables directly from your BI tool, without any compute cost: https://github.com/buremba/universql

Did Pinot replace Clickhouse at Uber or is this a different team/dataset?

Hi all! I built this tool for data exploration on Snowflake tables without any need to run a warehouse.

Snowflake is very powerful at scale but for small data (< 100GB) I rather work locally and then deploy my models to Snowflake later on. Universql lets you keep using Snowflake clients with its proxy layer and executes the queries locally on DuckDB.

This tech only became possible with Iceberg as it keeps the data references in manifest files where you can intelligently cache the data locally, similar to Snowflake's warehouse cache.

Interesting idea but this Etsy example didn't work for me :(

^ this. I had a similar feeling when I saw the landing page as I have been waiting for Sora to be available after it's "announced".

Does Vercel charge for the compute duration in addition to the number of requests for Next.js apps? I see they charge for Function Duration (https://vercel.com/docs/pricing/serverless-functions#managin...) but I am unsure if Next.js apps are considered as Vercel Functions.

^ Exactly my feelings about React. The sad truth is that React is still viable for startups as the ecosystem has the most extensive librarie.

Captcha seems to be the only solution to prevent it and yet this is the worst UX for people. The big publishers will probably get their cut no matter what but I’m not sure if AI will leave any room for small/medium publishers in the long run.

What's wrong with the URL in the query?

I can’t really think of a product with the requirement of max 6TiB data. If the data is big as TiB, most products have 100x TiB rather than a few ones.

Totally. They do that with PAIR nowadays.

^ this but it’s very hard to keep that editor open all the time if you’re not “obsessed” with what you are doing.

Since it's free, people probably don't think they are missing out as they can get that "free" at some point later on. You take things more seriously when you pay for them.

I believe Google can override its CPU budget manually though.

Congrats on the launch! How is the AI component better than using Copilot on VSCode?

It's the time for OpenAI to set a new bar with GPT-5 then. For reference, it's been more than a year since they released GPT-4 and Sam was indicating that the leap from GPT-4 to GPT-5 is going to be similar to how people felt GPT-3 -> GPT-4 here: https://www.youtube.com/watch?v=jvqFAi7vkBc&t=5274s

--you-know-what-im-talking-about

It would be amusing to be able to do:

<a href="bash:rm -rf /">Destroy me</a>

Neon.tech runs Postgresql runs on S3. They persist the WAL to S3 so that they can replicate the data and bring it to local ssds I assume.

That would probably require them to rewrite a non-trivial part of S3 from scratch.

I think even for SaaS products, the enterprise customer contracts should include statements around making the source/infra code available under an open license. That would help companies gain confidence in a way that they can always hire the ex-employees and continue using the product.

I don't think they call it "really necessary" but I assume it's handy to understand the user base and focus on the right use cases in the long run considering not all the users who use these libraries are active on Github.

Seems interesting as current Python-based solutions don't have good support for Javascript. I see the support for the binding is from Python to JS but do you have any plans to support two-way binding work?

It's not every that you run into a post on HN about the area you live in. :) After trying 5-6 Airbnb's across the city for almost a year, Angel seemed to be the best in terms of the mood, pubs, canal, shops, and restaurant. It's been years but I discover new spots around here quite often and probably stick to here for quite some time!